Risk screening solutions, such as those from LexisNexis, Thomson Reuters, and LSEG help bank compliance teams—especially those in financial crime compliance (FCC) operations—verify customer identities to comply with Know Your Customer (KYC) and Anti-Money Laundering (AML) regulations. These risk screening solutions each leverage their own proprietary global data.
Yet, compliance operations teams recognized long ago that none of these data providers supply a fulsome data set to their users. And so, for years, users have supplemented their screening solution data by searching for additional information using popular Web search engines, including Google, Bing, DuckDuckGo, and more—with Google being the most widely used by far.
Still, FCC leaders have continued to struggle to obtain all the data necessary to gain a clear picture of the people and entities that pose a risk to their banks.
But why do they continue to struggle? Doesn’t Google provide the critical mass of data, which when combined with the data from their screening solution, provides a full picture?
Short answer: No.
Google’s blind spots
The fact is that Google only indexes 15-20% of the public information that exists online, and the problem of incomplete information is rampant in Google search results. That’s because Google makes the final decision on what it will show a search user, and it does so based on many criteria, such as Google’s relevance scoring, the user’s location, and the user’s search history. Unfortunately for FCC teams, incomplete results hamper the information collection needed for KYC, AML, and fraud operations.
Also consider the information which becomes difficult to find due to SEO practices. SEO stands for ‘Search Engine Optimization’ which involves organizations making concerted efforts to promote their content over that of others. The content from organizations that do not excel at SEO tactics often ends up buried in Google search results (think page 10 instead of page 1).
Google also suffers from geographic blind spots. For example, according to the New York Times, Google’s core products are unavailable in China because of the country’s censorship restrictions. Back in 2010, Google pulled its search engine out of China due to government censorship and a cyberattack from Chinese hackers trying to gain access to human rights activists’ email accounts. For Google, China is not the only blind spot in the World. Similar censorship exists in countries like Iran, North Korea, Russia, Serbia, Turkey, and more.
Search results manipulated by states and other influential organizations
Perhaps worse than missing information is the presentation of disinformation. The fact is, Google is a revenue- and profit-driven business, and that can translate into inaccurate information gaining prominence in search results. For example, The Brookings Institution performed a 120-day study focused on COVID-19 and Xinjiang’s role in it. The study revealed that Chinese state media content featured prominently in 21.5% of the top results on Google News and Bing News.
That type of situation exists for Google searches across the globe and even here in the United States when influential or deep-pocketed organizations invest heavily in promoting the content they want users to see…or not see.
Extrapolate the Brookings findings across all the nations (including those in the EU and North America) that have a vested interest in shielding specific people and entities from public scrutiny, and you quickly realize why Google and the other Web search engines can provide both incomplete and inaccurate data to FCC analysts.
How leading banks are improving their FCC data & operations today
FCC leaders are working with solution providers that have years of AML, KYC, and related subject matter expertise to both gain more complete data and to improve how they make decisions on that data.
Large language models (LLMs) like ChatGPT, Anthropic, and Gemini do a great job of supplementing traditional Web search and other technology. But you still have to combine that with people who have domain and subject matter expertise to design the prompts and instructions which the LLMs will follow. When designed correctly, LLMs perform searches at scale, and the AI can then summarize all the articles and present a condensed version in written, understandable form to an analyst. And, of course, the information is footnoted. So, the user can, if they choose to do so, read the source article.
But robust results depend on subject matter domain expertise. For example, an analyst cannot simply open ChatGPT and prompt it with “give me negative news on Person X.” It doesn’t work that way, especially when a search must be performed in 40 different languages and work around the nuances which each individual nation or locality may present as roadblocks. And these roadblocks are very often NOT nefarious but just a matter of local focus or culture.
For example, certain financial jurisdictions (e.g., the Channel Islands like Jersey or Guernsey) have a tremendous amount of financial activity. It is a local specialty that receives coverage from many local news and information sources. There’s a high probability that Google and LLMs will not discover those sources unless a subject matter expert designs them into the search. Similarly, well-designed search strings and LLM prompts which subject matter experts design in languages such as Arabic, Cyrillic, and Mandarin yield far better results than do standard searches in English that ask Google and LLMs to translate, search, and return what they find.
Subject matter expertise also revolves around HOW financial crime risk events are reported. Some countries allow for openly disparaging commentary while others do not allow any. There’s a wide spectrum of how adverse media may or may not show in different places, and only a subject matter expert knows all the various terms that signal adverse commentary.
Leading banks are already doing it
Leading banks with increasingly sophisticated understanding of the nuances involved in gathering and analyzing AML, KYC, and fraud data are relying on subject matter experts and a wider variety of data sources to de-risk their operations.
Not only do they work with better data, but they are also getting the right answers faster. That’s because they do not just receive the best information, but they also receive it in a way that says, “here is all the relevant information, and here is the final answer.”
For example, let’s assume an adverse media search yields 100 articles. When the data provider is boosted with subject matter expertise, the system can automatically pair the results to just 19 articles which the system is sure about and one which it is not sure of. Rather than forcing the analyst to read 20 articles, the system summarizes (on its own) the 19 it is sure about in a well-written paragraph that explains what the articles say, along with the 19 footnotes, enabling the analyst to read them if desired. This all takes place whether the data is 20 years old or published within the last 12 hours.
Essentially, the system does all the work and presents the right answer to the analyst.
This is where FCC operations have progressed in leading-edge financial institutions. Are your FCC data and operations keeping pace?
Contact us to learn more: Request a Demo | WorkFusion